Comparison Between Tcp Reno and Tcp Tahoe Using Ns Simulator

NS2 Assignment #4 ECE6607 Fall 2010 Marion Le Borgne

Turn out : 1. 2. 3. 4. 5. Ns source code Scripts (awk, perl, gnuplot) Results Plots Brief explanations

A. No packet loss Answers to section A : 1. Plots General behavior of TCP Tahoe

Slow start Regions (exponentials)

Congestion avoidance Regions (lines)

2. General behavior of TCP Reno

Congestion Avoidance (strait lines)

Slow start regions (exponentials)

3. Slow-start and congestion avoidance regions are labelled by hand on the plots “cwnd VS time” Strait lines correspond to congestion avoidance regions Exponential parts corespond to slow-start regions. 4. What difference(s) do you observe between TCP Tahoe and TCP RENO congestion window? At first, there's no difference difference between TCP Tahoe and TCP Reno. Indeed, both Reno and Tahoe start by sending an Synsignal. Then the servers replies with an ACK. And Both Reno and Tahoe will go into slow start phase. When tcongestion occurs, TCP reno and tahoe begin to behave differently : Reno will go only one time into the slow start phase, contraty to tahoe that does it twice beforre beginin the congestion avoidance phase.

5. If you cut in half the bandwidth delay product (by reducing the delay from 100 to 50 ms) what happens to the maximum value of the cwnd (is is bigger or smaller)? The maximal value of cwnd is smaller. It is logical that maximum value of the congestion windows decreases since it is a gess of the maximum throughput. It the throughput decraeses, the congestions window has to be smaller. To get these results I used the following codes and scripts. PART 1 is for the general behaviour of TCP TAHOE, and PART 2 is for TCP RENO

Part 1: General behavior of TCP Tahoe
Codes and scripts of these part are made to generate plots of Sequence number VS time, Ack number VS time, Cwnd VS time. There are four steps : NS code, Awk scripts (to parse the trace file), Gnuplot scripts (ploting), and finally the runscript. Step 1 : NS...